TLDR: A University of Texas at Arlington (UTA) team, spearheaded by Professor Dereje Agonafer, is leading a Department of Energy (DOE)-funded project under the COOLERCHIPS initiative. Their goal is to drastically cut energy and water consumption in AI data centers by developing innovative hybrid cooling technologies, aiming to reduce cooling energy use from up to 40% to under 5%. This effort is crucial as AI infrastructure rapidly expands, particularly in Texas, which is becoming a major hub for data centers.
ARLINGTON, Texas – As the demand for artificial intelligence (AI) infrastructure surges across the nation, particularly in Texas, a critical challenge has emerged: the immense energy and water consumption required to cool the massive data centers powering generative AI. A pioneering team at the University of Texas at Arlington (UTA), under the leadership of Presidential Distinguished Professor Dereje Agonafer, is at the forefront of addressing this issue through a significant project funded by the U.S. Department of Energy (DOE).
This initiative is a cornerstone of the DOE’s COOLERCHIPS program, a $42 million endeavor launched in 2022. COOLERCHIPS, an acronym for ‘Cooling Operations Optimized for Leaps in Energy, Reliability, and Carbon Hyperefficiency for Information Processing Systems,’ aims to fundamentally rethink data center cooling. Professor Agonafer’s team is targeting an ambitious reduction in cooling energy use, striving to bring it down from as much as 40% of a data center’s total load to less than 5%. This effort also seeks to mitigate the substantial water usage associated with many large-scale data centers, which often rely on evaporative systems consuming hundreds of thousands of gallons daily.
‘The computing power needed to support AI is growing at an exponential rate,’ stated Professor Agonafer, a member of the National Academy of Engineering. ‘If we don’t find smarter, more efficient ways to cool data centers, the environmental impact will be significant. Our goal is to deliver a scalable solution that is both energy-efficient and environmentally responsible.’
Texas has become a pivotal location for large-scale data center development, driven by factors such as land availability, tax incentives, and robust energy infrastructure. However, this growth comes with considerable environmental and resource costs. The Electric Power Research Institute projects that U.S. data centers could consume up to 9% of the nation’s electricity by 2030. In Texas, the state’s main grid operator anticipates data center electricity demand to more than double, from 4.6% in 2023 to nearly 11% by 2030, according to UTA.
Water consumption is also a growing concern. Texas data centers are estimated to use 25 billion gallons of water in 2025, representing about 0.4% of the state’s total. This figure could escalate to 2.7% by 2030, equivalent to the annual water use of 1.3 million U.S. households, as reported by the Texas Tribune.
The UTA team is collaborating with researchers from the University of Maryland, the University of Illinois Urbana-Champaign, and the Illinois Institute of Technology, all focused on reducing energy and water use in increasingly dense server environments. The project also includes a crucial partnership with Accelsius, an Austin-based startup specializing in two-phase, direct-to-chip liquid cooling systems. Accelsius is contributing its MR250 Coolant Distribution Unit, designed for high-density server environments, a technology that received $24 million in Series A funding last year to expand its reach in the AI market.
Peter de Bock, ARPA-E program director, described COOLERCHIPS as a ‘moonshot program’ at the 2024 Data Center World conference, emphasizing the agency’s investment in immersion cooling, direct-to-chip approaches, and other advanced methods. De Bock anticipates a proof of concept by the first half of 2026, noting that ‘By 2030, the country with the most efficient, powerful, and lower-TCO data centers will be at a major advantage.’
Graduate students Sai Pundla and Braxton Smith are integral to Agonafer’s team, gaining invaluable hands-on experience in system-level testing and thermal management. ‘Being part of this project has been an incredible opportunity to apply what I’ve learned in the lab to a real-world challenge with global implications,’ said Smith.
The Dallas-Fort Worth area, ranked as the No. 2 data center market in North America by JLL research, is experiencing rapid development to meet surging demand. Several multi-billion dollar projects are underway, many specifically targeting high-density, high-performance AI workloads. These include Wistron Corp.’s $750 million AI supercomputing facilities in Fort Worth, PowerHouse and Provident’s planned 1.8-gigawatt campus in Grand Prairie, Stream Data Centers’ $300 million expansion in Wilmer, and Gigabit Fiber’s $1 billion communications infrastructure campus in Red Oak. Furthermore, OpenAI’s ambitious Project Stargate, in partnership with Oracle and SoftBank, is driving even greater scale, with five new campuses announced in September, alongside its flagship campus in Abilene, Texas, aiming for a $500 billion, 10-gigawatt commitment by the end of 2025.
Dan Diorio, vice president of state policy at the Data Center Coalition, highlighted the economic impact, stating, ‘Data centers are the backbone of the 21st-century economy, and they’re economic drivers for D-F-W and in Texas communities.’
Also Read:
- Generative AI’s Soaring Energy Demand: OpenAI’s Text-to-Video Models Could Rival India’s Power Consumption
- Meta’s Infrastructure Undergoes Major Evolution to Power AI Advancements
This comprehensive effort by UTA and its partners underscores a critical national push to ensure that the exponential growth of AI computing power is met with sustainable and efficient infrastructure solutions.
